It took me a couple of days to track this down.
I am using WebODM 1.9.18, Docker Desktop 4.15.0 (93002), on MacOS (Intel) Ventura. I use GCPEditor Pro to generate gcp_list.txt.
When I generate an orthophoto with no GCPs, it works fine. But when I add the gcp_list.txt to the list of files, I get the message “Cannot process dataset” about a minute into the task. I replicated this error on WebODM Lightning.
It turned out that the problem was that my JPEGs had spaces in the file names (e.g., 2022-09-26 15-24-47 DJI_0293.JPG). When I replaced the space with an underscore, the problem went away.
Clearly, this is in the code that reads the gcp_list.txt. It is probably using both spaces and tabs as field separators, even though GCP Editor Pro uses only tab as a separator character.